Updated
Claude Sonnet 5 vs GPT-5.5
GPT-5.5 had 78% fewer counted bugs per line than Claude Sonnet 5 (0.47 vs 2.10).
| Claude Sonnet 5 | GPT-5.5 | |
|---|---|---|
| Bug rate | ||
| Relative bug rate | 2.10 | 0.47 |
| Rank | #9 of 9 | #1 of 9 |
| Details | ||
| Provider | Anthropic | OpenAI |
| Released | Jun 30, 2026 | Apr 23, 2026 |
| List price | $2 input / $10 output per 1M tokens | $5 input / $30 output per 1M tokens |
| Share of AI-written code, week of Sep 14 | 4.0% | 0.1% |
1.00 = average · lower is better
More comparisons
Questions and answers
- Which has fewer bugs, Claude Sonnet 5 or GPT-5.5?
- GPT-5.5 had 78% fewer counted bugs per line than Claude Sonnet 5 (0.47 vs 2.10). GPT-5.5 ranks #1 of 9 and Claude Sonnet 5 #9.
- Which is cheaper, Claude Sonnet 5 or GPT-5.5?
- Claude Sonnet 5 lists at $2 input / $10 output per 1M tokens; GPT-5.5 lists at $5 input / $30 output per 1M tokens. Claude Sonnet 5 has the lower list price.
- Which is used more, Claude Sonnet 5 or GPT-5.5?
- Claude Sonnet 5, with 4.0% of the AI-written code reviewed by cubic in the week of Sep 14, 2026, against 0.1% for GPT-5.5.